“Those who cannot remember the past are condemned to repeat it” (Santayana, 1905). This warning is particularly relevant when studying artificial intelligence in the judicial process. Cases such as COMPAS show that automation is not neutral: an algorithmic system can reproduce human biases, conceal them under a technical guise, and undermine fundamental rights in high-stakes decisions (Guanche, 2023). Therefore, the central legal question is not whether AI can make the justice system more efficient, but rather what regulatory limits must be imposed to prevent efficiency from eroding due process, effective judicial protection, equality, privacy, judicial independence, and the right to a natural judge.
The comparative analysis must be limited to regulatory systems. According to Zweigert and Kötz’s functional method, one does not compare identical institutions, but rather systems that fulfill an equivalent function (Zweigert & Kötz, 1998). In this case, the common function of the European-Spanish model and the Mexican model is to regulate the use of AI in the administration of justice to protect human rights and prevent opaque, discriminatory, or unduly automated decisions.
The European system offers more comprehensive regulation. Regulation (EU) 2024/1689 adopts a risk-based model and classifies certain AI systems used by judicial authorities—or on their behalf—to interpret facts, interpret the law, or apply rules to specific cases as high-risk. This classification imposes enhanced obligations: risk management, data governance, technical documentation, record-keeping, transparency, human oversight, accuracy, robustness, and cybersecurity. In the judicial sphere, this means that AI cannot operate as a substitute for the judge or as an opaque source of decision-making.
Spain supplements this framework through Royal Decree-Law 6/2023, which distinguishes between automated, proactive, and assisted actions within the administration of justice. The distinction is essential: automated actions are reserved for simple procedures that do not require legal interpretation; assisted actions may generate drafts or complex documents, but these do not constitute judicial rulings without human validation and must be modifiable by the competent authority. The Spanish model, therefore, embraces technology while preserving the human core of the judicial system.
Mexico, on the other hand, still lacks a comprehensive law in force regarding judicial AI. However, its constitutional and treaty-based framework provides sufficient grounds for imposing limits. The Mexican Constitution recognizes access to justice, due process, the requirement to state reasons and provide justification, equality, property rights, the amparo proceeding, and the right to be tried by previously established courts. The Amparo Law requires special caution regarding any mechanism that could unduly restrict access to constitutional justice, as dismissal is only permissible in cases of manifest and indisputable grounds. Furthermore, the General Law on the Protection of Personal Data Held by Obligated Entities requires security, proportionality, and impact assessments when authorities process personal data on a large scale.
The proposed Mexican regulation, currently under discussion in the Senate, functionally aligns with the European model by incorporating transparency, auditing, traceability, protection of rights, human oversight, and control over technologies with significant impact. However, Mexico should not automatically adopt European regulations. It must adapt them to its constitutional tradition, the amparo proceeding, structural inequality, and the need to prevent systems trained on flawed historical data from amplifying discrimination.
The impact on human rights is central. AI can affect access to justice if it becomes a formalistic barrier to the admission of lawsuits or appeals. It can violate due process if the parties are unaware of which data, rules, or criteria influenced a ruling. It can compromise equality and non-discrimination if it reproduces biases found in case files, statistics, or precedents. It can affect privacy and the protection of personal data when processing financial, family, medical, criminal, or sensitive information. It can also erode judicial independence if a judge uncritically adopts automated recommendations.
Added to these risks is the impact on the right to a natural judge and the right to a human judge. The guarantee of the right to a natural judge not only requires a pre-established, competent, independent, and impartial court; it also requires that the decision come from a human authority capable of listening, contextualizing, weighing the evidence, and justifying the ruling. Tovar warns that the red line is drawn where technological assistance begins to replace human judgment, undermining independence, impartiality, reasoning, and effective judicial protection (Tovar, 2025). Therefore, an automated justice system may be formally jurisdictional, but in substance contrary to the right to be judged by a human judge.
The deeper risk is that poorly regulated AI may reinforce an outdated conception of the judge as a mere mechanical enforcer of the law, ignoring the fact that adjudication involves interpreting, arguing, evaluating, and deciding responsibly (Pantoja Morán, 2010). Furthermore, a judicial ruling does not merely apply general norms; it also tailors the law to individual cases and, to a certain extent, participates in its judicial creation (Bulygin, 2003). In conclusion, the European-Spanish model offers a preventive and rights-protecting approach; Mexico has a fragmented framework, but one that is constitutionally sufficient to demand limits. Mexican regulations should adopt a high-risk approach to judicial AI, incorporating transparency, explainability, traceability, auditing, data protection, meaningful human oversight, and a ban on fully automated decisions.
In conclusion, the European-Spanish model offers a preventive and rights-based approach to the use of AI in the judicial system, while Mexico still has a fragmented framework—albeit one that is constitutionally sufficient to impose limits. Any future Mexican regulation should not be based on uncritical trust in the technology or on an absolute ban on its use. The purpose of this analysis is not to argue that artificial intelligence should be excluded from the administration of justice, but rather to foster a critical, prudent, and constitutionally informed approach. AI can contribute to a more efficient, orderly, and accessible justice system; however, it can also pose risks and dangers that must be assessed, weighed, and regulated prior to its implementation. These risks are not solely technical, as they are not limited to programming errors, data failures, or cybersecurity issues. They are also legal, procedural, and democratic.
Poorly regulated judicial AI can affect access to justice, due process, equality, privacy, the protection of personal data, judicial independence, the reasoning behind rulings, and the right to a natural judge (Tovar, 2025). Furthermore, it can produce a form of justice that is seemingly objective but materially opaque, in which the parties do not know what data, criteria, or inferences influenced a decision. In such a scenario, automation would cease to be an auxiliary tool and would become an obstacle to effective judicial protection.
The most serious risk is that AI could silently transform the very conception of the judicial function. If used without caution, it could reinforce a formalistic view of the judge as a mere mechanical enforcer of the law, ignoring the fact that adjudication involves interpreting, weighing, evaluating evidence, contextualizing facts, and justifying decisions (Pantoja Morán, 2010). Furthermore, a judicial ruling is not merely an exercise in legal subsumption: it also individualizes the law and, within certain limits, participates in its judicial creation (Bulygin, 2003). Therefore, de facto delegating judicial reasoning to algorithmic systems would compromise not only specific procedural guarantees but also the very essence of the judiciary.
Consequently, Mexican regulations should adopt a high-risk approach to any use of AI in the judicial sphere. This requires transparency, explainability, traceability, auditing, human rights impact assessments, enhanced protection of personal data, meaningful human oversight, and a ban on fully automated decisions. AI can be useful for assisting, organizing, or accelerating certain aspects of the process, but its legitimacy will depend on its remaining subordinate to human deliberation. In the judicial function, the focus should not be on algorithmic efficiency, but on the individual, their rights, and the public responsibility of adjudication.
References
1. Bulygin, E. (2003). Los jueces ¿crean derecho? Isonomía, Revista de Teoría y Filosofía del Derecho, (18).
2. Cámara de Diputados del H. Congreso de la Unión. (2025). Ley de Amparo, Reglamentaria de los artículos 103 y 107 de la Constitución Política de los Estados Unidos Mexicanos. Diario Oficial de la Federación.
3. Cámara de Diputados del H. Congreso de la Unión. (2025). Ley General de Protección de Datos Personales en Posesión de Sujetos Obligados. Diario Oficial de la Federación.
4. Cámara de Diputados del H. Congreso de la Unión. (2026). Constitución Política de los Estados Unidos Mexicanos. Diario Oficial de la Federación.
5. Guanche, J. C. (2023). La historia del algoritmo. Los “fallos” de la inteligencia artificial. UNESCO.
6. Jefatura del Estado. (2023). Real Decreto-ley 6/2023, de 19 de diciembre. Boletín Oficial del Estado.
7. Pantoja Morán, D. (2010). Los jueces de la tradición: Un estudio de caso. Boletín Mexicano de Derecho Comparado, 43(127).
8. Parlamento Europeo y Consejo de la Unión Europea. (2024). Reglamento (UE) 2024/1689 por el que se establecen normas armonizadas en materia de inteligencia artificial. Diario Oficial de la Unión Europea.
9. Santayana, G. (1905). The life of reason: The phases of human progress. Charles Scribner’s Sons.
10. Senado de la República. (2025). Propuesta marco normativo. Comisión de Análisis, Seguimiento y Evaluación sobre la Aplicación y Desarrollo de la Inteligencia Artificial en México.
11. Tovar, M. (2025, 18 de noviembre). El derecho a un juez humano: la línea roja de la inteligencia artificial. Instituto de Democracia y Derechos Humanos de la Pontificia Universidad Católica del Perú.
12. Zweigert, K., & Kötz, H. (1998). An introduction to comparative law (3rd ed.). Oxford University Press.
About the Author
Juan José Herrera Zaballa, Jr. Associate at SOTMOR Consulting & Legal Management
He is a Mexican legal professional specializing in private law, civil and commercial matters, aviation law, amparo proceedings, litigation, and arbitration. As a Junior Associate at SOTMOR Legal Consulting & Management, he advises on contractual analysis, transactional matters, aviation projects, regulatory strategy, and complex dispute resolution. His international advocacy experience includes participation in the Leiden–Sarin International Air Law Moot Court Competition and the Willem C. Vis International Commercial Arbitration Moot. Building on a family legal tradition spanning more than six decades, he combines rigorous legal analysis with strategic thinking and an innovative approach to delivering practical, client-focused legal solutions.
